• Neuroinformatics
• Apollo close

Matrix factorisation treats observations as linear combinations of basis vectors together with, possibly, additive noise. Notable techniques in this family are Principal Components Analysis and Independent Components Analysis. Applied to brain images, matrix factorisation provides insight into the spatial and temporal structure of data. We improve on current practice with methods that unify different stages of analysis simultaneously for all subjects in a dataset, including dimension estimation and reduction. This results in uncertainty information being carried coherently through the analysis. A computationally efficient approach to correlated multivariate normal distributions is set out. This enables spatial smoothing during the inference of basis vectors, to a level determined by the data. Applied to neuroimaging, this reduces the need for blurring of the data during preprocessing. Orthogonality constraints on the basis are relaxed, allowing for overlapping ‘networks’ of activity. We consider a nonparametric matrix factorisation model inferred using Markov Chain Monte Carlo (MCMC). This approach incorporates dimensionality estimation into the infer- ence process. Novel parallelisation strategies for MCMC on repeated graphs are provided to expedite inference. In simulations, modelling correlation structure is seen to improve source separation where latent basis vectors are not orthogonal. The Cambridge Centre for Ageing and Neuroscience (Cam-CAN) project obtained fMRI data while subjects watched a short film, on 30 of whose recordings we demonstrate the approach. To conduct inference on larger datasets, we provide a fixed dimension Structured Matrix Factorisation (SMF) model, inferred through Variational Bayes (VB). By modelling the components as a mixture, more general distributions can be expressed. The VB approach scaled to 600 subjects from Cam-CAN, enabling a comparison to, and validation of, the main findings of an earlier analysis; notably that subjects’ responses to movie watching became less synchronised with age. We discuss differences in results obtained under the MCMC and VB inferred models. This work was supported by the Medical Research Council at the Biostatistics Unit [Unit Programme number U105292687].

Learning and experience are critical for translating ambiguous sensory information from our environments to perceptual decisions. Yet, evidence on how training molds the adult human brain remains controversial, as fMRI at standard resolution does not allow us to discern the finer-scale mechanisms that underlie sensory plasticity. Here, we combine ultra-high field (7T) functional imaging at sub-millimetre resolution with orientation discrimination training to interrogate experience-dependent plasticity across cortical depths that are known to support dissociable brain computations. Our results provide evidence for recurrent plasticity, by contrast to sensory encoding vs. feedback mechanisms. We demonstrate that learning alters orientation-specific representations in superficial rather than middle V1 layers, suggesting changes in read-out rather than input signals. Further, learning increases feedforward rather than feedback layer-to-layer connectivity in occipito-parietal regions, suggesting that sensory plasticity gates perceptual decisions. Our findings reveal finer-scale plasticity mechanisms that re-weight sensory signals to inform improved decisions, bridging the gap between micro- and macro- circuits of experience-dependent plasticity. $$ \ $$ See the file 'Description of uploaded data' for a detailed description of the dataset.

Multimodal biological and cognitive data used as predictors and outcomes for machine learning models can be found in 'master data sheet.xls'. With the exception of the derived PLS Derived GM all data were downloaded from the ADNI repository http://adni.loni.usc.edu/. For description on derivation of the PLS Dervived GM see ���Methods: Partial Least Squares Regression with Recursive Feature Elimination (PLSr-RFE).��� in the final publication DATA SETS: 1.) ���Methods: Partial Least Squares Regression with Recursive Feature Elimination (PLSr-RFE).��� Data available: RIDS: The ADNI identifier, DIAG(1CN, 2MCI): Baseline diagnosis (1:cognitively normal, 2: MCI) ADNI Mem: ADNI Memory composite measure used as outcome variable for the PLSr-RFE, PLS Derived GM: Variable derived from the PLSr-RFE procedure. These data are presented in ���Results: Composite grey matter score for predicting cross-modality associations��� 2.) ���Statistical Validation: Out-of-Sample validation[cross-modality associations ]��� Data available: RIDS: The ADNI identifier, DIAG(1CN, 2DEM, 3MCI): Baseline diagnosis (1:cognitively normal, 2:demented, 3: MCI), PLS Derived GM: Variable derived out-of-sample. FTP Braak(12): tau PET SUVR for Braak stage (1,2), FTP Braak(34): tau PET SUVR for Braak stage (3,4), FTP Braak(56): tau PET SUVR for Braak stage (5,6). These data are presented in ���Results: Composite grey matter score for predicting cross-modality associations��� 3.)���Statistical Validation: Out-of-Sample validation [Cross-modal associations -adni mem]��� Data available: RIDS: The ADNI identifier ADNI Mem: ADNI Memory composite measure used as outcome variable. These data are presented in ���Results: Composite grey matter score for predicting cross-modality associations��� 4.) ��� Methods:GMLVQ Cognitive model��� Data available: RIDS: The ADNI identifier, ADNI Mem: ADNI memory composite used as predictor, ADNI EF: ADNI executive function composite used as predictor, GDS: Geriatric Depression Score used as predictor. 1pMCI, 2sMCI: Outcome classes, 1:progressive Mild Cognitive Impairment, 2: stable Mild Cognitive Impairment. ���Results: Cognitive Classification Models for predicting sMCI vs pMCI��� 5.) ��� Methods:GMLVQ Biological model��� Data available: RIDS: The ADNI identifier, PLS Derived GM: grey matter score used as predictor, FBP: florbetapir SUVR used as a predictor, APOE4: APOE 4 genotype used as predictor. 1pMCI, 2sMCI: Outcome classes, 1:progressive Mild Cognitive Impairment, 2: stable Mild Cognitive Impairment. ���Results: Biological Classification Models for predicting sMCI vs pMCI��� 6.) ��� Methods: GMLVQ-Scalar Projection *Cognitive model*��� Data available: RIDS: The ADNI identifier, ADNI Mem: ADNI memory composite used as predictor, ADNI EF: ADNI executive function composite used as predictor, GDS: Geriatric Depression Score used as predictor, �� ADNI-Mem: Change in ADNI mem from baseline. 7.) ��� Methods: GMLVQ-Scalar Projection *Biological model*��� Data available: RIDS: The ADNI identifier, PLS Derived GM: grey matter score used as predictor, FBP: florbetapir SUVR used as a predictor, APOE4: APOE 4 genotype used as predictor, �� ADNI-Mem: Change in ADNI mem from baseline. ���Results: Trajectory modelling: Predicting Individual Variability in the Rate of Future Cognitive Decline. 8.) ���Methods: Statistical Validation: Out-of-Sample-[Cognitive model]��� Data available: RIDS: The ADNI identifier, ADNI Mem: ADNI memory composite used as predictor, ADNI EF: ADNI executive function composite used as predictor, GDS: Geriatric Depression Score used as predictor, �� ADNI-Mem: Change in ADNI mem from baseline. 9.) ���Methods: Statistical Validation: Out-of-Sample-[Biological model]��� : RIDS: The ADNI identifier, PLS Derived GM: grey matter score used as predictor, FBP: florbetapir SUVR used as a predictor, APOE4: APOE 4 genotype used as predictor, �� ADNI-Mem: Change in ADNI mem from baseline. ���Results: Trajectory modelling: Predicting Individual Variability in the Rate of Future Cognitive Decline.��� For a more detailed description of the populations these data were extracted for see 'description of uploaded files.doc'

SIGNIFICANCE: Functional near-infrared spectroscopy (fNIRS) is a frequently used neuroimaging tool to explore the developing brain, particularly in infancy, with studies spanning from birth to toddlerhood (0 to 2 years). We provide an overview of the challenges and opportunities that the developmental fNIRS field faces, after almost 25 years of research. AIM: We discuss the most recent advances in fNIRS brain imaging with infants and outlines the trends and perspectives that will likely influence progress in the field in the near future. APPROACH: We discuss recent progress and future challenges in various areas and applications of developmental fNIRS from methodological and technological innovations to data processing and statistical approaches. RESULTS AND CONCLUSIONS: The major trends identified include uses of fNIRS "in the wild," such as global health contexts, home and community testing, and hyperscanning; advances in hardware, such as wearable technology; assessment of individual variation and developmental trajectories particularly while embedded in studies examining other environmental, health, and context specific factors and longitudinal designs; statistical advances including resting-state network and connectivity, machine learning and reproducibility, and collaborative studies. Standardization and larger studies have been, and will likely continue to be, a major goal in the field, and new data analysis techniques, statistical methods, and collaborative cross-site projects are emerging.

Background Mapping the hypoxic brain in acute ischemic stroke has considerable potential for both diagnosis and treatment monitoring. PET using (18)F-fluoro-misonidazole (FMISO) is the reference method; however, it lacks clinical accessibility and involves radiation exposure. MR-based T2' mapping may identify tissue hypoxia and holds clinical potential. However, its validation against FMISO imaging is lacking. Here we implemented back-to-back FMISO-PET and T2' MR in rodents subjected to acute middle cerebral artery occlusion. For direct clinical relevance, regions of interest delineating reduced T2' signal areas were manually drawn. Methods Wistar rats were subjected to filament middle cerebral artery occlusion, immediately followed by intravenous FMISO injection. Multi-echo T2 and T2* sequences were acquired twice during FMISO brain uptake, interleaved with diffusion-weighted imaging. Perfusion-weighted MR was also acquired whenever feasible. Immediately following MR, PET data reflecting the history of FMISO brain uptake during MR acquisition were acquired. T2' maps were generated voxel-wise from T2 and T2*. Two raters independently drew T2' lesion regions of interest. FMISO uptake and perfusion data were obtained within T2' consensus regions of interest, and their overlap with the automatically generated FMISO lesion and apparent diffusion coefficient lesion regions of interest was computed. Results As predicted, consensus T2' lesion regions of interest exhibited high FMISO uptake as well as substantial overlap with the FMISO lesion and significant hypoperfusion, but only small overlap with the apparent diffusion coefficient lesion. Overlap of the T2' lesion regions of interest between the two raters was ∼50%. Conclusions This study provides formal validation of T2' to map non-core hypoxic tissue in acute stroke. T2' lesion delineation reproducibility was suboptimal, reflecting unclear lesion borders.

Behavioural data and DTI connectivity data (see supporting data description .doc file for more information)

Perception is the process by which our brains interpret sensory information. Our brains are constantly evaluating sensory signals in our environments, which shapes how we experience the world and, ultimately, our physical and mental well-being. When this develops differently, it may lead to atypical sensory perception as seen in autism. The addition of sensory symptoms to the most recent diagnostic criteria for autism highlights the need to understand its underlying mechanisms. This thesis used methods from experimental psychology and brain imaging to investigate the neurocognitive mechanisms of perceptual inference in autistic individuals. Chapter 1 introduces the topic of sensory perception, its neurocomputational framework, and its role in autism. It provides an overview of the theories and models of perception in autism and presents the overarching aims of this research. Chapter 2 reports a study of how autistic adults make perceptual decisions on two visual similarity judgment tasks. Signal detection theory analyses indicated that, in both tasks, when compared to typical people, autistic individuals used different decision criteria during conditions of uncertainty. Chapter 3 addresses the limited neuroimaging research on non-social features of autism. Using activation likelihood estimation, findings were condensed from non-social perception task-based functional MRI studies examining differences between autistic and typical participants. Overall, autistic people, compared to typical controls, showed less activity in the prefrontal cortex during perception tasks. More refined analyses revealed that, when compared to typical controls, autistic people showed greater recruitment of the extrastriate cortex during visual processing. Chapters 4 and 5 report findings from a visuomotor probabilistic reversal learning task used to examine how adults with varying levels of autistic traits evaluate sensory information, build, and update sensory expectations. A positive relationship was found between autistic traits and the learning of probable sequences before the reversal. In addition, there were separate main effects of autistic traits and intolerance to uncertainty on the ability to update expectations following the reversal. These findings suggest that, while people with different levels of autistic traits identify statistical regularities at a comparable level to one another, autistic traits play a role in how individuals update their expectations once a change is introduced. Chapter 5 examined how these behavioural findings relate to inhibitory neurotransmitters. In this 7-Tesla MR spectroscopic investigation, γ-Aminobutyric acid (GABA) was measured in the occipital and motor cortices to investigate its role in visuomotor sequential learning and its interactions with autistic traits. Previous findings of a negative relationship between sensorimotor GABA and sequence learning were replicated. At the same time, there were no clear links between autistic traits and occipital and motor GABA. Finally, Chapter 6 ties these findings together and evaluates how they contribute to our understanding of autistic perception. Some of the challenges of cognitive neuroscience research in autism are highlighted alongside clear directions for future work.

Identifying targets in cluttered scenes is critical for our interactions in complex environments. Our visual system is challenged to both detect elusive targets that we may want to avoid or chase and discriminate between targets that are highly similar. These tasks require our visual system to become an expert at detecting distinctive features that help us differentiate between indistinguishable targets. As the human brain is trained on this type of visual tasks, we observe changes in its function that correspond to improved performance. We use functional brain imaging, to measure learning-dependent modulations of brain activation and investigate the processes that mediate functional brain plasticity. I propose that dissociable brain mechanisms are engaged when detecting targets in clutter vs. discriminating between highly similar targets: for the former, background clutter needs to be suppressed for the target to be recognised, whereas for the latter, neurons are tuned to respond to fine differences. Although GABAergic inhibition is known to suppress redundant neuronal populations and tune neuronal representations, its role in visual learning remains largely unexplored. Here, I propose that GABAergic inhibition plays an important role in visual plasticity through training on these tasks. The purpose of my PhD is to investigate the inhibitory mechanisms that mediate visual perceptual learning; in particular, learning to detect patterns in visual clutter and discriminate between highly similar patterns. I show that BOLD signals as measured by functional Magnetic Resonance Imaging (fMRI) do not differentiate between the two proposed mechanisms. In contrast, Magnetic Resonance Spectroscopy (MRS) provides strong evidence for the distinct involvement of GABAergic inhibition in visual plasticity. Further, my findings show GABA changes during the time-course of learning providing evidence for a distinct role of GABA in learning-dependent plasticity across different brain regions involved in visual learning. Finally, I test the causal link between inhibitory contributions and visual plasticity using a brain stimulation intervention that perturbs the excitation-inhibition balance in the visual cortex and facilitates learning.